JPS6324478A - Generation of 2-d data display for indicating 3-d data set - Google Patents

Abstract

(57) [Abstract] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of the Invention The present invention relates to a method for generating two-dimensional data representations representing three-dimensional data sets.

[Conventional technology]

In a wide range of modern applications, it is desirable to observe the three-dimensional coherence of a volume or object of interest. When generating images of real three-dimensional (3D) solid bodies, it is originally only possible to view individual planar cross-sections of the 3D volume and its contents. It is typically not possible to view a 3D volume in a visually discernible manner by separating the interior surfaces, surfaces of objects, boundaries, boundaries, interfaces, and spatial relationships within the volume.

For example, in video generation in the medical field, it is highly desirable to be able to visualize anatomical structures through a three-dimensional representation on a computer graphics display. The ability to generate accurate, voluminous anatomical models from computerized tomographic (CT) scans makes it possible to develop accurate, voluminous anatomical models in surgical practice (to plan surgical procedures, without the need to conduct ridges or pilot surgeries). It is extremely valuable as a display (such as displaying a structure to show the details of a structural structure). therefore,
The 3D shape, dimensions and relative positions of pathological structures provide important data for surgical treatment planning, diagnosis and treatment.

Computer simulations of real 3D volumes rely on the ability to reconstruct 3D structures from planar cross-sectional data such as CT scans. These and other scans yield data that allows a 3D density image volume consisting of voxels to be used as input data for image processing. Unfortunately, such input video volumes typically have low resolution compared to the level of detail desired to accurately represent the extracted volume.

〔problem〕

For example, in a scanned image volume at C, voxels represent x-ray attenuation or other image volume data throughout the volume, including boundaries across the surface.

Although a voxel is assigned one "uniform" value, there is boundary material and discrete material on either side of the boundary within the object under consideration. Thus, a voxel along an edge is a specimen that extends across both sides of that edge. Furthermore, if an object (such as a bone) is thinner than one voxel wide, the voxel resolution that provides boundary information about the bone is very low. Therefore, the shape of the boundaries within a voxel is not easily determined.

Various techniques have been used to approximate surface boundaries within a volume. One well-known technique is the "thresholding" method. In the thresholding method, voxels across a boundary are classified as being composed of class a of one material type or another material type on either side of the boundary. Therefore, the projected visible boundary becomes a binary classified voxel boundary.

The larger the voxel, the greater the error introduced by the thresholding technique. Furthermore, for coarse images that map dense, closely spaced surface boundaries, thresholding techniques yield even worse results; becomes increasingly inaccurate. Later approximation techniques are sometimes used to obtain a more accurate approximation of the results obtained by the thresholding technique. However, attempts to approximate unknown surfaces produce highly inaccurate results. The reason is that their approximations depend on the initial binary classification of the voxels.

It is also highly desirable to be able to simultaneously view all data within a volume from a selected static or rotating view. That is, it is highly desirable to see the center of a volume and to detect objects and hidden surfaces (and thus internal boundaries and interior surfaces) within the volume. To this end, it is necessary to be able to see partially through the obstructing object when desired (for example, in the case of a bone surrounded by muscles, the surrounding muscles as well as the It is necessary to be able to observe the Conventional techniques display volume elements that allow a binary decision to be made as to whether a certain pixel is made of a given material. Because binary classification does not preserve continuous video functions, binary classification introduces additional (ie, sampling) errors. Errors caused by binary classification occur on any attempt to reconstruct the original video volume from the revealed volume. Since the reconstructed image can only have as many bright reflection levels as there are materials present, the material interfaces are jagged and the bright reflection represents the original image function;

[Summary of the invention]

With the imaging device of the present invention, surface boundaries and objects inside the volume are easily shown, and hidden surfaces and surface boundaries themselves are accurately displayed. ) provides an apparatus and method for projecting representations. In addition, the generated two-dimensional image can have the sampling resolution and gardening resolution of the target input image volume. This is done by an implementation method that determines the "partial volume" of the voxel. Partial volume determination allows the selected color and opacity (
opacity) to the various materials (i.e., data components) represented in the video data volume based on the percentage of material represented within each volume of the video volume. Unlike conventional devices that use techniques to determine the image quality, the image generation technique of the present invention is highly accurate and provides high definition that is independent of the image volume per voxel.

A video volume representing a volume object or data structure is written to an image storage device. A color and opacity are assigned to each voxel in the volume, red (R), green (G),
Blue (B) and opacity (A) components, stored as a three-dimensional data volume. RGB for each volume
The A assignment is based on the percent component composition of the materials coated within the volume, and therefore the percent color and transparency l' associated with those materials (the color value and transparency of the 100% reference material). Such voxels that are stored and arrayed in the image storage device for the M, minute channel volume are also referred to herein as RGBA voxels.

The voxels within the RGBA volume are then used as a mathematical "gel" filter, such that the voxel filter is superimposed on top of the conventional background voxel filter.

In addition to linear interpolation, a new background filter is determined and generated. The interpolation is performed successively for all voxels (or groups of voxels, e.g. volume scan lines) up to the most anterior voxel with respect to the plane of the drawing.

This method is repeated until all visible voxels have been determined for the plane of the view.

The reconstruction method of the present invention allows υ to reconstruct the surfaces of objects within individual data volumes, even in cases where representations could not previously be seen due to coarse video volume data or low sampling violations. It provides a two-dimensional projection of a volume so that details can be extracted and displayed. The present invention visualizes hidden and hidden surfaces within a volume so that all surface boundaries are clearly and precisely defined.

Therefore, given a complex volumetric data set, i.e. an object, it is possible to calculate the percentage of the composition of the constituent materials for each voxel without first being given specific information about the location of the boundaries within each volume. It is possible to display complex images. Thereby, the method and apparatus of the invention utilizes partial transparency and color that can have the characteristics of a geometric solid model.
It is possible to create ffi projections of 3D volumes. Objects that partially obscure other objects can be seen and spatial relationships between objects can be accurately displayed.

['A Example]

Hereinafter, the present invention will be described in detail with reference to the drawings.

The detailed description that follows is primarily of algorithms and symbolic representations of operations on data pits within a computer storage device. Algorithmic description and representation are the means by which those skilled in the data processing arts most effectively convey their Q Fit to others in the field.

An algorithm is understood here as a series of self-contained processes that lead to a desired result. Those processes are those that require physical handling of physical ft. Usually, but not necessarily, these quantities take the form of electrical or magnetic signals capable of being stored, transmitted, combined, compared, and otherwise manipulated. It has proven convenient at times, principally for reasons of common convenience, to refer to these signals as bits, heads, elements, symbols, characters, terms, numbers, or the like. However, all these and similar terms should relate to appropriate physical quantities,
It should be remembered that it is just a label attached to one of them.

Furthermore, the operations performed may also refer to things like additions or comparisons that are commonly associated with mental tasks performed by an operator. However, in any of the operations described herein that constitute the present invention, it is not necessary or desirable in most cases for an operator to be able to do so. Those operations are machine-executed operations. An Isosei device useful in carrying out the operations of the present invention is a general purpose digital computer or similar device.

In all cases, it should be noted the difference between the way computers operate and how they operate, and the way the computer itself calculates. The present invention relates to a method of operating a computer in the processing of electrical or other physical (e.g. mechanical or chemical) signals to generate other desired physical signals.

The invention also relates to a method of performing those operations. The apparatus may be specifically configured for the required purpose, or it may include a general purpose computer that can be selectively activated or reconfigured by a program contained within the computer. The algorithms described herein are not inherently related to any particular computer or other apparatus. In particular, it has been found convenient to use a variety of general-purpose machines or even more specialized equipment to carry out the required method steps.

Although the term voxel is sometimes used for convenience to refer to a three-dimensional element of a three-dimensional volume 44, it should be understood that a voxel defines a point in three-dimensional space.

In the following description, numerous details are set forth regarding algorithmic conventions, specific numbers of bits, etc., in order to provide a thorough understanding of the invention.

However, it will be apparent to one skilled in the art that the present invention may be practiced without such specific details. In other instances, well-known circuits and structures have not been described in order to avoid obscuring the present invention in unnecessary detail.

System Architecture The system architecture of the present invention is shown in FIG. In the preferred embodiment of the present invention, the video system of the present invention hosts a host computer 10 coupled to a bitmap terminal 12 and a device bus 14. device bus 14
is a channel processor 16, a storage device controller 18,
A video controller 20 is coupled to the video controller 20 . Storage device controller 18
is coupled to channel processor 16, video controller 20 and image storage [26]. Video controller 20 is coupled to image storage 26 via video bus 28. Channel processor 16 is coupled to image storage 26 via image bus 24. A color monitor 22 or other output device is coupled to the video controller 20.

The channel processor 16 is a parallel processing computer device. In the detailed described embodiment of the invention, channel processor 16 employs four 16-bit parallel processors. In the present invention, it has been found that it is advantageous to use four parallel processors, but
It will be apparent that any number of processors may be used without departing from the spirit of the invention.

Bitmap terminal 12 displays a number of menus available to a user of the invention. In the embodiment described herein, host computer 10 is used to manipulate histograms generated from image data stored in image storage device 26. A lookup table can be stored in the host computer 10 that includes color values and opacity values assigned to each peak 4 degree location in the histogram.

The host computer 10 can also be used to display a histogram on the bitmap terminal 12 so that the user can make down payments on defined color values and opacity. Although a 16-bit parallel processor is used in the present invention, 8-bit, 32-bit, or any other size processor could be used. Therefore, device buses of other pit sizes can be used as required.

Video controller 20 provides vertical and horizontal synchronization to color monitor 22 and also includes pan storage for displaying screen data.

Video controller 20 includes a digital-to-analog (D/A) converter that converts the digital signal to a video signal suitable for display on monitor 22. Video controller 22 also includes lookup tables for each RGB channel. For example, a 12-pit digital word in the red (R) channel is mapped through a lookup table and a unique red value is output and displayed on color monitor 22. As mentioned above, each channel R,
Search tables are provided in G and B. Video controller 20 functions as a ternary input-to-ternary output converter. The three-value input
Gamma correction is performed using a three-value output converter. In the embodiment described herein, video bus 28 is a bus that is divisible into 12 pit units and operates at least 420 megabytes per second.

In this embodiment, the image bus 24 is dimmed 24 for one second.
It runs on 0 megabytes.

The color monitor 22 of the present invention can be any commercially available color monitor having 480 to 1125 scan lines. The present invention has the speed and performance to operate high resolution and multiple scan line monitors.

The image storage device 26 includes random access memory (RGB) used to store original video data, classification data, color, opacity data, and composite RGBA volume data.
RAM). Image storage device 26 is shown in greater detail in FIG. As mentioned above, each voxel consists of four 12
Represented by Bit Digital Go. Image storage device 26
The original volume storage 26(a), which is a part of the , stores data in its original form. In the embodiment described herein, approximately 40 million bits are dedicated to storing raw data. The original video volume data is stored as a series of 12 pit words. Each of the 12 pit words represents the brightness level of one voxel of the original voxel volume. After the RGBA volume is generated, the RGBA volume of Sochi et al. is stored in the RGBA volume storage device 2.
6(b). Each color value is represented by 12 pit words, and the opacity A is also represented by 12 pit words. Therefore, each voxel in the RGBA volume storage is represented by four 12-pit words. In the detailed described embodiment of the invention, 160 million bits of storage are dedicated to RGBA volume storage 26(b). Composite voxels are stored in a two-dimensional projection storage device 26 (d). As mentioned above, combined voxels are generated from the concatenation of RGBA volume data.

Image storage 26 is temporary workspace storage 12
6(C) is also included. Although the number of bits in the various portions of image storage 26 have been described, it will be apparent to those skilled in the art that any suitable number of bits of storage may be utilized to implement the present invention.

Referring again to FIG. 1, storage controller 18 is used to arbitrate all accesses to image storage 26. Storage controller 18 enables data to be written to image storage 26 via image bus 24 .

A channel processor 16 is used to generate histograms used to perform voxel classification. The channel processor 16 is shown in detail in Figure @3. This channel processor includes a scratch pad memory 17,
41 multiplier/arithmetic logic unit (ALU) 15a-15
d. In the embodiment described herein, scratch pad memory 17 includes 64 16-bit word storage locations. The channel processor 16 includes a multiplier/
A scratchpad memory 17 is used to temporarily store data handled by the ALU pair. Each multiplication 5/ALU
Pairs 15a to 15d are the four pairs forming the channel processor.
dedicated to one of the processors. This parallel technique allows the same instruction to be executed on η different pieces of data. The system is referred to in the art as a single instruction, multiple data stream (SIMD) system. Such a one-instruction, multi-data flow system is developed by the applicant.
It is described in detail in U.S. Pat. No. 748,409, dated June 24, 1985. In the example described here, one channel is the red (R) of the voxel.
one channel is used for the voxel's green (G) value, and one channel is used for the voxel's 'fl?' value. (B
) values and one channel is used for the voxel opacity red (A) values.

When the original video volume data is read into the host computer 10, initial settings of the processing device are performed. Since the original video volume data can be sent in compressed form, it requires decoding and decompression.

Also, depending on the source of the data, the bits/bits of raw data
The high cell resolution may not match the preferred 12 bits/pixel resolution of the present invention.

Therefore, if desired, the host computer or channel processor can increase the resolution of the original video data to an acceptable resolution.

Host computer 10 then outputs the original video data to channel processor 16 via device bus 14. Channel processor 16 then writes the original video data to original video volume storage 26 of image storage 16 via image bus 24 .

If desired or necessary, the original video volume is classified by channel processor 16. Scan lines from the original volume store @26 are input to the channel processor 16 via the image bus 24. The channel processor 16 generates a histogram by counting the number of voxels at each brightness level of the 2049 possible NL levels. Channel processor 16 can then classify the peaks according to a previously defined lookup table or program hierarchy. The histogram can also be displayed on the pit map terminal 12 by the host computer 10 so that the classification can be performed manually by the user.

After classification, the classification data is sent via device bus 14 to channel processor 16, in particular to scrubber memory 17. The channel processor calculates a plurality of lookup tables based on classification data from the host computer. Scratchbud memory 17 is
It includes multiple search clothes so that RGB@ and opacity values can be assigned to the classified voxels. Channel processor 16 requests one scan line (one dimensional array of voxels) at a time from the storage controller and processes the scan lines through a stored lookup table. The input to the channel processor search cloth is a black and white scanning line, and the output is RG.
This is a scanning line classified as BA. Channel processor 16
The scratch pad memory 17 includes at least three buffers for output classification: a scan line buffer, a lookup table buffer, and an output buffer for output classification. Although described in connection with lookup tables, any suitable method of assigning or generating color values may be utilized in the present invention.

The output from the output buffer is transferred via the image bus 24 to the RGBA volume storage device 26 of the image storage device e26 (b)
is combined with Each slice of the original video volume is stored as a play in RGBA volume storage 26). The “concatenated filter” of the present invention
ecl filtering) J is executed by the channel processor 16. The background image volume is
One scan line at a time is initially read into the channel processor's scratch pad memory 17. As each successive video volume is read into the channel processor 16, the RGBA values are concatenated by multipliers/ALUiSa~15d.

This composite video volume is stored in projection storage 26(d) of image storage 26.

The contents of projection storage 26(d) are output via video bus 28 to video controller 20 for displaying the composite image. The digital data stored in the projection storage device 26(d) is converted into video analog data by the video controller 20 and then output to the color monitor 22, where it is displayed in a rask scan manner.

After a partial volume classification video volume, i.e. a volume of image elements (voxels) representing a three-dimensional video, has been read into the host computer 10 and any required decoding or decompression has been performed, the original data in the image storage 26 is Video storage device 26 (
b). Therefore, the object considered here, ie an array of two two-dimensional arrays, can be thought of as a three-dimensional ordered array in an image storage device, and can be represented mathematically. The video volume can be an unprocessed video array that can be obtained in a variety of known ways.

Although the invention has wide applications, it will be described by way of example in connection with three-dimensional display of computerized tomography (CT) images. Input image volume data, in this case represented as an ordered CT image volume array, is obtained from various tomographic image generation methods, such as X-ray computerized tomography, ultrasound sector scanning, nuclear magnetic resonance, etc. be able to. In other cases, seismic imaging or computer models or (fluid flow simulation)
Other image generation methods can be used to obtain the input image volume data, such as the results of a simulation. The video volume in this example is stored as an ordered array of two-dimensional images, each image being a two-dimensional ordered play of a 12-bit number.

A CT constant scan or other image volume provides monochrome grayscale input data to the image processing system of the present invention. This C
T input video volume data is stored in volume store 26(a)K of image store 26 as an ordered array of 12-bit binary numbers. Each binary number represents a given CT constant scan or other image volume data (
voxel) table.

In the example described here, the image volume provided by the CT scan input consists of four individual materials in the anatomy under consideration: air, fat,
It is provided by a CT scan input representing information about soft tissue and bone.

In the embodiment described herein, the intensity of the grayscale input data depends on the x-ray intensity of the material presented from the original image generation method. Referring to FIG. 4a, the gray-scale intensity data for each voxel is plotted in a histogram 30 giving the distribution of the number of voxels in the image volume versus intensity.

This histogram 30 is generated by the channel processor 16 on the device bus 14 which is composed of input video volume data.
The data is sent to the host computer 10 via the host computer 10.

The gray tone histogram 30 is a function that indicates, for each gray tone, the number of voxels in the video volume that have a particular gray level. The horizontal axis 32 is gray level intensity, shown in this example as O~2048. The vertical axis 34 is the frequency of occurrence (eg, the number of voxels in the image volume at each intensity). Therefore, the area under the histogram curve 36 (111) represents the total number of voxels in the image volume.

Histogram 30 is obtained from a particular image corresponding to CT constant scan 12-bit pixel information in the embodiment described herein. The resulting histogram is filtered by the host computer 10 using low-pass filtering techniques widely referred to in the art to remove noise, i.e., while preserving the shape of the human function.
This results in a smoothed histogram curve as shown in FIG. 4a.

In a conventional video generator bus, all voxels are classified as representing 100% of a single material. Therefore, using conventional techniques, a 2M classification is performed for each voxel. In the binary classification, all voxels within a particular intensity region are classified as one or the other of those materials represented by the input video volume data. Using conventional techniques for this example CT scan, the volumes would be classified as bone, fat, or air. In the prior art, a threshold gray level is selected as a cutoff at a point at the peak of the histogram. All voxels within a given intensity range are thereby classified as 100% air, soft tissue, bone or fat. This information is stored in storage as a 2-bit value or a 1-bit value.

For CT display, this binary classification is appropriate for the peak region of the histogram. In that peak region, the contents of the voxel fall into one individual material class or another individual material class. However, more than that, “...-de”
Classification requires making individual binary decisions regarding the material classification of all volumes within the video volume.

For example, in conventional techniques, at a surface boundary (such as tissue attached to bone), voxels that cross the edge region are classified as bone or tissue. Therefore, when a voxel crosses between surfaces, or across an edge, or if the local material is smaller than one voxel wide, the edge or surface boundary is lost, resulting in inaccurate display. The value assignment introduces large additional errors, resulting in inaccurate surface representations.

In the present invention, voxels are classified according to their associated intensities as consisting of 0-100% of each material type represented in the image volume.・Next, refer to the iJb diagram. Big beak P and big valley T
is shown in the histogram 30.

Strength range 40 marked as 100% of the material
is determined for all voxels falling within the selected intensity range to the left or right of the large histogram peak P. In the embodiment described here, in the case of CT constant scanning, the intensity range is determined by linear approximation. 100% material coverage is approximated as all voxels P falling within the intensity band defined as 1p-tl/4. where p is the intensity data related to the large peak,
t is the intensity associated with the adjacent large valley T. All voxels within this range on either side of a given large beak P are classified as 100% "pure" material values. Therefore, in this example, all voxels within those ranges are classified as 100% fat, air, soft tissue, and bone, respectively. Voxels outside those intensity ranges are classified as containing some video information representing a certain percentage of "pure" material on the left or right side of the voxel. They are therefore combinatorial material voxels, each having a partial percentage composition of those materials.

For convention purposes, voxels classified as being composed of less than 100% of any single material are referred to as partial voxels in the context of the present invention.

Additionally, voxels classified as 100% of one material are sometimes referred to as such. However, in more general terms using the present invention, classification is based on partial volume determinations. Therefore, as a more general convention, every voxel classified is a "partial volume" υ, and a voxel classified as one uniform material contains 100% of one material and no other component materials. is 0%
The composition is classified as a "partial volume".

In determining the classification of the subvolumes, the percentage of each material contained in voxels outside the 100 material g1 degree range is found. In the example described here, the percentage voxel material content is determined by performing linear interpolation on voxels whose intensities fall between the limits of adjacent 100% material intensity values. . Then, (1) the intensity difference (numerator) from the 100% value boundary to the target voxel strong #, and (2) the 10
The ratio between the limits of the 0% material value and the strength range value (denominator) is found.

This ratio gives the percentage voxel composition for each adjacent material at a particular intensity value between the 9.100% value range.

Reference is now made to Figure 4b. The partial volume percentage is set to 100 adjacent 100 voxels for every partial volume voxel.
% determined by channel processor calculations performed by linear interpolation for each strength between the material strength value limits n
Ru. Alternatively, it can be done according to the values listed in the lookup table, or 451! This will be done according to the first shift determined by the meal and the user. In the example described here, for a given voxel associated with a partial volume [I(pv), its percentage gain is given by:

Part of the material in the voxel PV3b Here, PVab is the target partial volume voxel, ■(P■) ab is the intensity of the target partial volume voxel between the 100% material strong reversal regions a and b, I(Pb )
& is the intensity at the peak boundary value for 100% material type, and Pb is the intensity at the boundary value for 100% material type.

When calculating the part of the material related to boundary a, I(Pb
)a is used and I(Pb) is used when calculating the part of the material associated with the boundary.

Therefore, each voxel consists of (1) a subvolume made up of percentages of two types of video volume material, and (2
) "Pure" composed of one type of video volume material
Classified as a volume.

In the embodiment described herein, the classification of grayscale data is performed on a musculoskeletal CT scan model. In this model, as shown in Figure 4b, the relationship between the gray torso strength of the material is such that fat 40 is always adjacent to soft tissue 44, soft tissue 44 always surrounds fat 40, and bone 46. Always surround. However, it should be understood that other methods of classification may be useful for determining the percentage voxel material F) composition for other input data, such as, for example, seismic volumetric input data. Such classification can also be used in applications where the input data includes material adjacencies. In that material contiguity, many materials (! or their associated histogram intensities, or both) can be simultaneously adjacent to each other and thus represented by the video volume input data.

Additionally, although this example describes a linear interpolation method, the orthogonal or cubic methods, for example, can be used to calculate voxel fractional volume percentages in this and other examples.

In the present invention, a 2D color image of a 3D volume is constructed based on a color representation of the image volume using "concatenated filtering" as described below in the reconstruction section.

To create such a filter, a color value (RGB) and an opacity value are randomly assigned to each voxel of the video volume. Once all the voxels in the volume have been classified and the percentage material composition of those voxels is known, a color (RGB) and opacity (A) can be assigned.

In this example of a color model generation technique for CT image volume data, bone is opaque white, soft tissue is translucent red, fat is very transparent green, and air is completely transparent.

A value having red (R), blue (B) and green (G) components is assigned. The values represent the color values of each of those four substances. Furthermore, an opacity (or transparency) is assigned to the selected transparency of each 100c6 pure material.

Knowing the RGBA assigned to each type of material, the RGBA can be determined for each voxel in the image volume based on its percent composition. For example, if a voxel is classified as 30% air, 70% fat, the "combined" RGBA determined by the channel processor 16 is added to the RGBA of each material component in the voxel with its associated percent material contribution. By multiplying by and adding the results, we can find for voxels with and . The resulting RGBA
is the RGBA of the target voxel, which is R
The data is read into the GBA volume storage device 26 (b).

therefore, becomes. Here, W = contribution of material i in the voxel, C=color and opacity of material i.

and, Here, n is ≠ of the component material in the voxel.

The RGBA voxel allocation operation is performed for each voxel in the video volume of interest, and their values are R
GBA volume image storage 26 (written to bc and stored for later processing. Each RGBA
A-image storage 26(b) is represented as a particular number of ordered arrays.

Each is an associated voxel of the video volume.

In conventional video generation devices, each voxel is typically assigned only one type of material and/or color value. All voxels in the volume were therefore considered to be uniformly composed of one type or another of the material of the video volume. In contrast, in the present invention, a wide range of colors and opacity is determined and assigned based on color and opacity combinations that represent the associated macrofractional material contribution to each voxel. The RGBA allocation in the present invention is based on R, G in the scrub pad memory 17 of the channel processor 16.
, B, and A by processing the classified voxels through the search tables.

Once the image regrooving RGBA volume has been created and stored in the RGBA image storage device, a 2D projection can be constructed for any selected view of the 3D volume. In the 2D projection, objects that are not represented as mathematical solids or surfaces of the 3D volume give the visible properties of the 2D projection made from the model of the mathematical surface or solid. This is done using the concatenated filters of the present invention.

Refer now to FIG. A local ordered 3×3 array of display pixels 60 is shown in this figure. The 3X3 display and 3X3X5 volume 62 shown in FIG. 6 is used for the example illustrating the present invention. In practice, the display is a 1024X 7684 or other suitable raster scan display. Furthermore, the RGBA volume is a three-space data array representing the (X + yr Z ) video volume, constrained only by the input data and the size of the storage device. Each RGBA voxel acts as a color filter to create a combined 2D output as shown in FIG. 5 for video display pixel (2,1).

The RGBA voxel as a filter can be viewed as a J voxel collapsed by 1° along zm (f, the axis parallel to the line of sight). ) gives a mathematical overlay. Each filter has a color and an opacity. Here, color is represented by three components: red (R), green (CG), and f (B). Opacity is represented by filter component (A). The opacity (A) of a filter is a measure of how many voxels behind the filter "show through" (i.e.
(contribution to new, combined background elements).

As shown in FIGS. 5 and 6, in order to provide a direct image on the XY plane, the RGBA of each voxel decoy (for example, 72) in the foreground is are each mathematically superimposed on the RGBAO of background voxels BG (e.g. 70) to create a new background RGBA
Generate voxel BG'. RG related to voxel BG
The BA values are read from the image storage device and temporarily stored in a scratchpad memory which acts as a BG filter buffer. Similarly, RG related to voxel BG
The BA value is read from the image storage device and stored in scratchpad memory. A part of that memory operates as an FG buffer.

In this embodiment, the color projected through the array of two or more concatenated filters is determined by the following equation.

BG""FG"(1(FG))X(Ayc))(BG
) where: FG = foreground filter color and opacity component G1AFG
= opacity of the foreground filter, BG = composite color and opacity G of all filters behind the foreground filter (ie, the "current" background).

Here, FG is the opacity of the foreground filter, i.e.
It represents the opacity contribution of the foreground filter to the new background filter BG. The resulting new background value BG'
is stored in scratchpad memory in preparation for calculations performed by the channel processor.

In this way, a new background filter BG is determined.

The above is then repeated for each foreground filter and its associated new background BGK until all filters are extended to the selected projection plane 64 and the final RGBA projected pixel constant is determined for the display 60. continues to be calculated.

This operation is performed to determine all display pixels for the selected 2D plane 64 of the image volume 62 of interest.

The reconstruction method of the present invention is not constrained by any particular processing operation, such as the pixel-by-pixel processing just described. For example, in this embodiment, successive voxel scans of an RGBA volume (i.e., a row of voxels perpendicular to the 2D plane 60 along a horizontal slice 66 of the RGBA volume 62) stored in the image storage device 26(b) Video reconstruction processing is performed on the image. Therefore, the RGBA of each foreground scan line BG is sequentially mathematically
Each background scan @BG is "overlaid" to generate a new background scan line BG'. R associated with scan line BG
The GBA value is read from the image storage device and temporarily stored in scratchpad memory.

Similarly, the RGBA values associated with the scan IBc are read from the image storage and stored in the scratchpad memo IJK-temporarily. A series of filter overlay calculations are
Subsequent associated BG and FG filters are read from the scratchpad memory and executed by the schannel processor to form the RGBA composite scan line 68.

That scan line is then written to video display output 64. Additionally, those operations are performed on selected 2D views of the 3D volume of interest to determine all composite pixel projection display scanlines.

Channel processor multiplier/ALU unit 1
5a-15d, filtering is performed by combining successive video volume parts (one scan line at a time). One channel (multiplier/ALU unit) is used to filter the R value. G value, B value and A
Each l channel is used to filter the values, so one channel is used for each voxel's RGB value. Filtering can be done from back to front or from front to back. The terms "front" and "back" are based on the viewing plane.

For example, the tables and equations shown at the end include the projected display inclination pixel 2 shown in Figures 5' and 6.
, 1, the white coat for RGBA voxels 70-78 is shown operating as a continuous concatenated filter. As can be seen from this table, the RGBA components of each value of each filter are calculated as a new BG' filter (then the filter continues to be <FG
) and the last vixel color (2,
1,). Therefore, subsequent calculations are performed according to the concatenated filter formula to calculate the new background filter frli to follow. These new background filter values are combined according to the above formula to calculate the draw value for each new foreground until the final pixel color value is determined. This value is then displayed (in this case by its associated display pixel (2,1)).

The composite 2D projection is then stored in display buffer 26 (d) and output to raster scan monitor 22.

Of course, it should be noted that the composite 2D projection can also be output to a printer or other output device.

In this way, a two-dimensional color projection of the three-dimensional input image volume is generated. Other image processing such as rotation, translation, shading, stretching, etc. can also be used based on the resulting projection or intermediate volume data.

The present invention can be used in a wide range of video processing. for example,
Voxel-sampled volumes (especially CT scans),
In particular, it can be used to display any data set that can be modeled in three dimensions, such as solution sets or calculation results from equations or other data sets that are not easy to model due to their complexity. Furthermore, the invention can be applied to color datasets where the colors of the component materials in the volume are known. Therefore, another means of sorting the data for percentage material contribution can be used to perform partial volume determination in accordance with the present invention.

Therefore, the present invention is not limited to grayscale datasets, cT constant scan real images, etc. Rather, the present invention
For each specimen (or voxel), a determinable individual volume specimen such that the type of material in the data cut (or differentiable enough to represent or assign the type of material) is known or can be found. Since the present invention can be applied to any voxel in the data volume, the percentage contribution can be found for each voxel in the data volume.

Although this example used a histogram distribution to perform subvolume percentage classification for CT constant scans, the classification can be performed for other applications without the use of histograms (i.e., spatially rather than statistically). This can be done in a variety of ways.

The technology for generating a two-dimensional display of a three-dimensional video volume has been described above.

(Tables and formulas) Below the filter components connected to each other, E: D+, 75Fjl; = (,22,,
08,,07,,55) DE under C: C ten, xo
g = (-31:S, , 55..28..955) CDE under B: s + , 8CDE = (,40,, 5B
, ,24. ．． 96) BCDE under A: A+, 7
Bcog=(,31,,65,,23,,97)

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the architecture of the present invention, FIG. 2 is a block diagram showing the division of the image storage device of the present invention,
FIG. 3 is a block diagram illustrating the architecture of the channel processor of the present invention; FIG. 4a is a graph of a typical input data histogram; FIG. 4b is a graph illustrating the components of a CT input data histogram; FIG. FIG. 6 is a diagram illustrating a plurality of voxels used to generate a display; FIG. 6 is a diagram illustrating subsequent voxel scan lines and image reconstruction of the workspace; 10φ---Host computer, 12@Φ book/bitmap terminal device, 14... Device lotus, 16...
Channel processor, 18...Storage device controller, 2
0...Video controller, 22-...Color monitor,
24... Image bus, 26...° Image storage device, 2
8...Video bus.

Claims (5)

[Claims] (1) generating a first video volume representing a three-dimensional data set; and generating the first video volume as a plurality of uniform storage arrays each comprising a plurality of scan lines having a plurality of volume elements (voxels); generating a second video volume by assigning a color value and an opacity value to each volume in a first video storage device; generating a combined image volume by combining said scans into lines; and performing said process for each successive said scan line with each generated combined image volume, where n is the total number of said scan lines. repeating up to the nth scan line, the combined image volume is configured such that the color value and opacity value of the voxel included in the combined scan line is A method for generating a two-dimensional data representation representing a three-dimensional data set characterized by having voxels based on a percentage composition of values. (2) The method according to claim 1, wherein the first video volume is converted into a plurality of digital signals and stored in a digital storage device. (3) The method according to claim 1, wherein the color value includes a red component, a green component, and a blue component. 4. The method of claim 3, wherein the combined video volume is generated by a channel processor, and wherein the red component, the blue component, the green component, and the opacity value are , a second channel, a third channel and a fourth channel, respectively. (5) The method according to claim 1, which
= color and opacity values of voxels in one said scanline, A = opacity of said voxels in one said scanline, BG = composite of corresponding voxels in previously combined scanlines A method characterized in that, as color and opacity components, the color and opacity values of each voxel of the combined video volume are given by FG+(1-(FG)(A))(BG).